Water Quality AI System

A full-stack water quality monitoring system: sensor readings flow in over an HTTP ingest endpoint, a Python service runs anomaly detection and short-horizon predictions, and a React dashboard surfaces the predictions, the raw signal, and an alert feed.

The frontend is React 18 with Recharts for the time series and a custom alert panel. The backend has two services: an Express 5 + PostgreSQL service for the ingest endpoint and the user-facing REST API, and a Python / Flask service that owns the model and the forecasting job. The two services share a single PostgreSQL instance; the model service writes its forecasts and the API service reads them.

The whole stack is Dockerised — docker compose up brings up the three containers (web, api, model) and the database. The model is a small LSTM; the training script ships in the repo and reads from a CSV fixture so the pipeline is reproducible without real sensor hardware.

Open the repo